UBA Ubiquitin Receptor Proteins Bind a Common Tetraubiquitin Chain

doi:10.1016/j.jmb.2005.12.001

J. Mol. Biol. (2006) 356, 1027–1035

UBL/UBA Ubiquitin Receptor Proteins Bind a Common Tetraubiquitin Chain Yang Kang1,2, Rebecca A. Vossler1, Laura A. Diaz-Martinez3 Nathan S. Winter4, Duncan J. Clarke3 and Kylie J. Walters1* 1

Department of Biochemistry Molecular Biology and Biophysics, University of Minnesota, Minneapolis, MN 55455, USA 2

Department of Oral Sciences University of Minnesota, Minneapolis, MN 55455 USA 3

Department of Genetics Cell Biology and Development University of Minnesota, Minneapolis, MN 55455 USA 4

Department of Chemistry St. Cloud State University St. Cloud, Minnesota MN 56301, USA

The ubiquitin–proteasome pathway is essential throughout the life cycle of a cell. This system employs an astounding number of proteins to ubiquitylate and to deliver protein substrates to the proteasome for their degradation. At the heart of this process is the large and growing family of ubiquitin receptor proteins. Within this family is an intensely studied group that contains both ubiquitin-like (UBL) and ubiquitin-associated (UBA) domains: Rad23, Ddi1 and Dsk2. Although UBL/UBA family members are reported to regulate the degradation of other proteins, their individual roles in ubiquitin-mediated protein degradation has proven difficult to resolve due to their overlapping functional roles and interaction with each other and other ubiquitin family members. Here, we use a combination of NMR spectroscopy and molecular biology to reveal that Rad23 and Ddi1 interact with each other by using UBL/UBA domain interactions in a manner that does not preclude their interaction with ubiquitin. We demonstrate that UBL/UBA proteins can bind a common tetraubiquitin molecule and thereby provide strong evidence for a model in which chains adopt an opened structure to bind multiple receptor proteins. Altogether our results suggest a mechanism through which UBL/UBA proteins could protect chains from premature de-ubiquitylation and unnecessary elongation during their transit to the proteasome. q 2005 Elsevier Ltd. All rights reserved.

*Corresponding author

Keywords: Rad23; Ddi1; ubiquitin receptor proteins; proteasome-mediated protein degradation; ubiquitin-associated domains

Introduction Ubiquitin signaling regulates an astounding array of cellular events and remains essential throughout the life cycle of a cell. In its most established role ubiquitylation targets proteins for degradation by the 26 S proteasome,1 a process important for controlling the lifespan of regulatory proteins, removing misfolded proteins,2 producing immunocompetent peptides, 3 activating and repressing transcription,4,5 and regulating cell cycle progression.6 In addition ubiquitylation can signal proteasome-independent events including Abbreviations used: Rad23, radiation-sensitive mutant 23; Ddi1, DNA damage-inducible protein; hHR23, human homolog of Rad23; HSQC, heteronuclear single quantum coherence; R4B, Rad4 binding domain; UBA, ubiquitinassociated; UBL, ubiquitin-like; GST, glutathione-Stransferase. E-mail address of the corresponding author: [email protected]

endocytic sorting7,8 and DNA repair.9,10 Ubiquitylation is connected to proteasome-mediated protein degradation by an intricate network of ubiquitin recognition proteins. Elucidating this network remains a difficult albeit active area of research, as it is clouded by redundancy and cooperation between the large and growing ubiquitin receptor protein family. Among these proteins exists a group that harbors both ubiquitin-associated (UBA) and ubiquitin-like (UBL) domains (Figure 1(a)). UBL/UBA proteins have attracted much attention for their ability to regulate the lifespans of other proteins. In Saccharomyces cerevisiae, Rad23 (hHR23a/b in humans), Dsk2 (hPLIC-1/2 in humans) and Ddi1 are UBL/UBA proteins that recruit ubiquitylated substrates to the proteasome for their degradation11–16 via UBA domain interactions with ubiquitin17–19 and UBL domain interactions with the proteasome.20–23 Depending on their protein levels, UBL/UBA-containing proteins can also inhibit the degradation of ubiquitylated substrates.16 Such inhibition occurs because UBA

0022-2836/$ - see front matter q 2005 Elsevier Ltd. All rights reserved.

1028

UBL/UBA Proteins Bind a Common Tetraubiquitin

UBL

(a) Rad23 1

UBA1 146 186

77

Ddi1

252

R4B 313

UBL 1

UBA2 355 395 UBA 389 428

75

(b)

105

110

110

115

15N

115

120 120 125 125 Black: Rad23 Red: Rad23 + Ddi1

10

9

8 1H (ppm)

Black: Ddi1 Red: Ddi1 + Rad23

130 7

6

(ppm)

(c) 105

10

9

1H

8 (ppm)

domains sequester K48-linked polyubiquitin chains to in turn prevent their elongation and de-ubiquitylation.16,24,25 In a perhaps related role, C-terminal UBA domains are reported to protect Rad23/hHR23a, Ddi1 and Dsk2 from their own degradation via the proteasome.26 Adding to the complexity of the ubiquitin family network, Rad23 is reported to interact with Ddi1 and Dsk2, and yeast two-hybrid experiments implicate UBA/UBA domain interactions as essential for such dimerization.27,28 However, in previous work on hHR23a we found no such UBA/UBA domain interactions but instead found that the UBL domain of hHR23a interacts dynamically with each of its UBA domains.29 Furthermore, whereas Rad23 is reported to dimerize,18,27,28 hHR23a does not.29 To resolve these ambiguities we used NMR spectroscopy to determine the mechanism by which Rad23 binds Ddi1 and itself. Yeast twohybrid experiments are unable to discriminate between direct interactions and those that are mediated by other proteins, both of which, as we report here, are available to Rad23 and Ddi1. Here, we reveal that UBL/UBA and not UBA/UBA domain interactions result in heterodimerization of Rad23 and Ddi1. These findings demonstrate the first published example for UBL/UBA domain interactions mediating heterodimerization. According to its crystal structure, the ubiquitin moieties of K48-linked tetraubiquitin are packed against each other with only the most distal moiety available for binding a UBA domain. 30 This structure suggests that K48-linked tetraubiquitin, which is the smallest chain length that signals for proteasome degradation31 is able to bind only one

130

7

Figure 1. Ubiquitin recognition proteins Rad23 and Ddi1 interact. (a) The sequence location of Rad23’s and Ddi1’s UBL and UBA domains as well as Rad23’s Rad4/ XPC binding domain is illustrated. (b) Comparison of the [1H,15N] HSQC spectrum of 15N-labeled Rad23 alone (black) to that acquired in the presence of fourfold molar excess Ddi1 reveals that Ddi1 causes severe broadening of certain Rad23 amide resonances (boxed in black). (c) [ 1H,15N] HSQC spectra are displayed of Ddi1 alone (red) and with fourfold molar excess Rad23 (red). The zoomed region highlights chemical shift perturbations in Ddi1’s [1H,15N] HSQC spectrum caused by Rad23 addition. Together (b) and (c) provide strong evidence for the direct interaction of Rad23 with Ddi1.

ubiquitin receptor protein. In addition, Rad23 has two UBA domains, the C-terminal of which is reported to sandwich between the two ubiquitin subunits of diubiquitin.32 Surprisingly, we have found that K48-linked tetraubiquitin can bind simultaneously to two Rad23 molecules as well as to Rad23 and Ddi1. This finding illustrates that K48-linked tetraubiquitin adopts an opened structure when bound to its receptors and leads to a working model for how ubiquitylated substrates are transferred to the proteasome.

Results Rad23 heterodimerizes with Ddi1 via UBL/UBA domain contacts Yeast two-hybrid experiments suggest that Rad23 interacts with itself, Dsk2 and Ddi1 by using UBA/ UBA domain interactions.27,28 These in vivo analyses, however, do not exclude the possibility of indirect associations through bridging molecules such as ubiquitin chains, which are known to interact with UBA domains.18 To test whether Rad23 and Ddi1 interact in their purified forms we performed [1H,15N] heteronuclear single quantum coherence (HSQC) experiments on 15N-labeled Rad23 or Ddi1 alone and in the presence of the other protein (Figure 1(b) and (c)). Such experiments detect amide nitrogen and proton atoms and their frequencies in spectra depend on their chemical environment; a phenomenon that makes them useful for detecting protein–protein interactions. The [1H,15N] HSQC spectrum of Rad23 resembles

1029

UBL/UBA Proteins Bind a Common Tetraubiquitin

that of its human homolog hHR23a, in that many sharp resonances appear in the region expected for randomly coiled residues (Figure 1(b)). This attribute originates from their long unstructured flexible linker regions that connect each of their small domains,29 which for Rad23 comprise 42.7% of its amino acid residues. In each set of [1H,15N] HSQC experiments direct interaction between Rad23 and Ddi1 was detected by spectral changes (Figure 1(b) and (c)). For Ddi1 the changes were confined to chemical shift perturbations (Figure 1(c)), whereas resonances derived from Rad23 experienced severe broadening and signal decay (Figure 1(b)). Although structural data and chemical shift assignments are available for the 40 kDa hHR23a protein,29 neither Rad23 nor Ddi1 has been characterized structurally by NMR or X-ray crystallography. Therefore, to identify the domains involved in forming the Rad23/Ddi1 protein complex we performed [1H,15N] HSQC experiments on single domain constructs of Rad23’s UBL, UBA1 and UBA2 domains and on Ddi1’s UBA domain. Comparisons between the truncated and full-length protein constructs allowed us to identify resonances derived from Rad23’s UBL and UBA1 domains (Figure 2(a)) as well as Ddi1’s UBA domain (data not shown). These resonance assignments were then used to interpret the results of the titration experiments recorded on the full-length proteins (Figure 1(b) and (c)). At equimolar protein concentration Rad23’s UBL domain interacts with Ddi1’s UBA domain and additional contacts are made between Ddi1’s UBL domain and Rad23’s internal UBA domain when Rad23 is present at twofold molar excess or greater (data now shown).

To test these UBL/UBA domain interactions more directly we performed analogous experiments in which we titrated the UBL domain of Rad23 into 15 N labeled full-length Ddi1 (Figure 3(a)) or its UBA domain (Figure 3(b)). We confirmed that the UBL domain of Rad23 binds the UBA domain of Ddi1 (Figure 3(b)), and found that the UBL domain of Rad23 induces identical chemical shift changes in Ddi1’s [1H,15N] HSQC spectrum compared to equimolar concentrations of full-length Rad23 (Figure 3(a)). Finally, we tested directly whether either UBA domain of Rad23 can bind that of Ddi1 by performing [1H,15N] HSQC experiments on 15Nlabeled Ddi1 UBA domain alone and in the presence of either of Rad23’s UBA domains. Even at eightfold molar excess Rad23’s UBA domains do not cause chemical shift changes in that of Ddi1. Therefore, we conclude that Rad23 UBA domains do not interact with that of Ddi1 (data not shown). Rad23 and Ddi1 dissociate to bind ubiquitin Since Rad23 and Ddi1 bind ubiquitin with their UBA domains, we tested whether the Rad23/Ddi1 heterodimer remains intact in the presence of monoubiquitin. To 15N-labeled Rad23 mixed with fourfold molar excess Ddi1 (Figure 4(a)), we added monoubiquitin such that the molar ratio of Rad23/ Ddi1/monoubiquitin is equal to 1:4:10 (Figure 4(b)). At 1:4:10 molar ratio, the resonances derived from Rad23’s UBL domain are restored, indicating that it is no longer bound to Ddi1’s UBA domain. This loss of interaction is due to monoubiquitin successfully competing for Ddi1’s UBA domain. In addition, resonances of Rad23’s UBA1 and UBA2 domains shift upon ubiquitin addition, confirming that each

(b)

(a)

(c) 105

115

15N (ppm)

110

120

125 Black: Rad23 Red: Rad23UBA1 Blue: Rad23UBL

10

9

8

7

6

Black: Rad23UBL Red: Rad23UBL+Rad23UBA1

10

9

8

7

10

Black: Rad23UBA1 Red: Rad23UBA1+Rad23UBL

9

8

130

7

1H (ppm)

Figure 2. Rad23’s UBA1 binds its own UBL domain. (a) The [1H,15N] HSQC spectrum of 15N-labeled Rad23 (black) is superimposed onto that derived from single domain constructs of its UBA1 (red) and UBL (blue) domains. (b) Superimposed [1H,15N] HSQC spectra of 15N-labeled Rad23’s UBL domain alone (black) and at equimolar concentration with its UBA1 domain (red). (c) The converse experiment is shown: [1H,15N] HSQC spectra of 15N-labeled Rad23’s UBA1 domain alone (black) and with its UBL domain at equimolar ratio (red). Selected shifted UBL and UBA1 resonances are boxed by green rectangles and ovals, respectively.

1030

UBL/UBA Proteins Bind a Common Tetraubiquitin

(a)

(b) 105

105

110

110

15

115

N (ppm)

115

120 120 125

125 130 Black: Ddi1 Red: Ddi1+Rad23 Blue: Ddi1+Rad23 UBL

10

9

8

Black: Rad23 UBL Red: Rad23 UBL+Ddi1 UBA

10

7

1

9

8 1

H (ppm)

7

H (ppm)

Figure 3. Rad23/Ddi1 heterodimerize via UBL/UBA domain interactions. (a) The [1H,15N] HSQC spectrum of 15Nlabeled Ddi1 alone (black) and with fourfold molar excess of either full-length Rad23 (red) or its UBL domain (blue). Rad23 and its UBL domain induce identical shifts in Ddi1’s UBA domain. (b) The [1H,15N] HSQC spectrum of 15Nlabeled Rad23 UBL domain alone (black) and with Ddi1’s UBA domain at equimolar concentration (red). Selected resonances of Rad23’s UBL domain that shift by adding Ddi1’s UBA domain are boxed.

of Rad23’s UBA (Figure 4(b)).

domains binds

ubiquitin

Rad23 forms a homodimer through its C-terminal half In addition to forming a heterodimer, Rad23 and Ddi1 each homodimerize. Ddi1 dimerization is independent of its UBL and UBA domain and occurs through residues located in the middle of its amino acid sequence.27 These residues are absent from the [1H,15N] HSQC spectrum acquired on fulllength Ddi1. Such absences are caused by line broadening due to chemical exchange or slow tumbling times (from bulkiness) and offer further support of an internal Ddi1 dimerization domain. That Ddi1’s UBL and UBA domains are observable reflects their structural independence from its dimerization domain. [1H,15N] HSQC spectra reveal that the resonances of Ddi1’s UBA domain superimpose well onto those derived from its full-length protein, as only two residues experience chemical shift changes due to this truncation (data not shown). These results support a model in which Ddi1’s UBA and UBL domains are autonomous and connected to the rest of the protein by flexible linker regions.

In contrast to Ddi1, the mechanism by which Rad23 dimerizes is not well understood. Dynamic light-scattering experiments indicate that 99% of Rad23 species (at 23.6 mM) exist as a dimer in its purified form (data not shown). Interestingly, its human homolog hHR23a does not dimerize29 and Rad23 dimerization was hypothesized to occur via UBA/UBA domain interactions.27 In our [1H,15N] HSQC experiments chemical shift changes compared to full-length Rad23 were observed for its UBA1 domain when produced as a single domain construct (Figure 2(a)). Such changes suggest interactions with other regions of the protein and we performed titration experiments with single domain constructs to discover that it binds its own UBL domain (Figure 2(b) and (c)). Most surprising, however, was that most of the resonances from the XPC/Rad4-binding and UBA2 domains were absent from the [1H,15N] HSQC spectrum recorded on full-length Rad23. These absences persisted even in an experiment performed for 14 h at 800 MHz with 128 increments in the 15N dimension and 128 scans per increment on a 15N-labeled 0.5 mM Rad23 sample. In contrast, all resonances derived from the XPC-binding and UBA2 domains are prominent in spectra recorded on hHR23a, even in experiments recorded with

1031

UBL/UBA Proteins Bind a Common Tetraubiquitin

(a)

(b) 105

Black: Rad23 Red: Rad23+Ddi1+Ub

Black: Rad23 Red: Rad23+Ddi1

15

115

N (ppm)

110

120

125

130 10

9

8

7

6 1

10

9

8

7

6

H (ppm)

Figure 4. Rad23 and Ddi1 dissociate to bind ubiquitin. (a) [1H,15N] HSQC spectra of 15N-labeled Rad23 alone (black) and with fourfold molar excess of Ddi1 (red) reveals that resonances of Rad23’s UBL domain disappear upon binding Ddi1. (b) [1H,15N] HSQC spectra of 15N-labeled Rad23 alone (black) and with Ddi1 and ubiquitin at 1:4:10 molar ratio, respectively, reveal ubiquitin to restore the broadened Rad23 resonances (boxed in black), and to cause chemical shift perturbations in Rad23’s UBA1 (boxed in blue) or UBA2 (boxed in green) domains. Certain UBA2 resonances appear upon Ddi1 or ubiquitin addition (marked with an asterisk).

eight scans per increment and 0.1 mM sample concentration.29 Interestingly, certain UBA2 resonances of Rad23 appear upon addition of Ddi1 (Figure 4(a)) or ubiquitin (Figure 4(b)); however, those of the Rad4-binding domain remain absent. These data support previous findings that Rad23 homodimerizes18,27 and suggest a role for the Rad4binding domain, which is consistent with the published finding that UBA2 is not sufficient for Rad23 dimerization.28 The N-terminal half of Rad23 including its UBL and UBA1 domain does not appear to be in contact with its C-terminal half and they are most likely connected by a flexible linker region that allows them to move independently of one another, as the N-terminal but not the C-terminal half of Rad23 is observable. Tetraubiquitin bridges ubiquitin receptor proteins We were interested in how the ability of Rad23 and Ddi1 to form homo- and heterodimers impacts their polyubiquitin-binding mechanisms and in particular whether one tetraubiquitin molecule can bind more than one ubiquitin receptor protein. We therefore tested whether tetraubiquitin can bind more than one Rad23 or Ddi1 molecule. To test

whether tetraubiquitin is capable of binding more than one Rad23 molecule we incubated K48-linked tetraubiquitin with Ni-NTA agarose resin containing pre-bound His-Rad23. After removing the unbound tetraubiquitin, glutathione-S-transferase (GST)-Rad23 was added and the beads washed again to remove unbound GST-Rad23. The resinbound protein complex was fractionated by gel electrophoresis, transferred to a membrane, and probed with anti-GST or anti-ubiquitin antibody (Figure 5(a)). This experiment revealed the presence of a ternary complex containing GST-Rad23, HisRad23 and tetraubiquitin. Since ubiquitin binding is reported to cause Rad23 homodimers to dissociate,27 we hypothesized that this ternary complex must be formed by each Rad23 construct binding directly to tetraubiquitin. In addition, this experiment was performed at a temperature (4 8C) that does not permit monomer exchange in Rad23 homodimers as GST-Rad23 and His-Rad23 failed to interact in the absence of tetraubiquitin (Figure 5(a), lower panel, lane 4). To test whether its human homolog hHR23a occupies all ubiquitin moieties of tetraubiquitin we examined whether the hHR23a–tetraubiquitin complex can also bind Rad23 (Figure 5(b) and (c)). Indeed, in experiments analogous to that described

1032

4

Ub

His-Rad23+GST-Rad23

Anti-Ub

GS T-h HR GS 23 a+ T+ Ub Ub 4 GS 4 T-h HR 23 a+ Ra d

(b) Ub4

-R

His

His

Ub

4

-R

ad

(a)

ad

23

+U

23

b4 23 His UB -R L+ ad Ub 23 4 +G ST -R ad 23

UBL/UBA Proteins Bind a Common Tetraubiquitin

Rad23

GST-hHR23a+Ub4

Anti-Ub Anti-Rad23

Anti-GST 2

3

4

5

6

1

7

2

3

4

5

6

7

-R

Ub

ad

4

23

9

+G di1

23

His

-D

ad -R

8

ST

L+ UB

b4 +U His

GST-hHR23a

His-Rad23+Ub4

His

4

-D

di1

(d) Ub

Ub

(c)

4 His -R ad 23 His +U b4 -R ad 2 3U His BL -R +U ad 23 b4 +G ST -hH R2

3a

1

1

2

3

4

GST-Rad23

His-Ddi1+Ub4

Anti-Ub

Anti-Ub Anti-GST

Anti-GST 1

2

3

4

5

6

7

8

9

5

6

7

Figure 5. Ubiquitin receptor proteins Rad23 and Ddi1 bind a common tetraubiquitin molecule. In lanes 5–7 of (a), 20 ml of Ni-NTA resin pre-incubated with 0.1 nmol of His-Rad23 was mixed with 0.1 nmol of GST-Rad23 and increasing quantities of K48-linked tetraubiquitin, 0.008, 0.033, 0.132 nmol, respectively, and probed with anti-ubiquitin (top panel) and anti-GST (bottom panel). As a negative control, this experiment was performed on His-Rad23 UBL mixed with tetraubiquitin (lane 3), or in the absence of tetraubiquitin (lane 4). As a positive control, the experiment was performed without GST-Rad23 (lane 2). The 0.01 nmol of K48-linked tetraubiquitin (Boston Biochem.) used in these experiments and those in (b), (c) and (d) was loaded directly onto lane 1 of each gel. In lanes 5–9 of (b), 20 ml of glutathione SSepharose resin pre-incubated with 0.1 nmol of GST-hHR23a was mixed for 1 h at 4 8C with 0.033 nmol of tetraubiquitin and increasing quantities of untagged Rad23: 0.033, 0.067, 0.1, 0.2 and 0.4 nmol, respectively. After washing the resin the proteins were fractionated, transferred to a membrane and probed with anti-ubiquitin (top panel) or anti-Rad23 (lower panel). Indeed, the tetraubiquitin-Rad23-GST-hHR23a protein complex was observed. In lanes 5–9 of (c) the converse experiment was performed whereby 20 ml of Ni-NTA resin pre-incubated with 0.1 nmol of His-Rad23 was mixed for 1 h at 4 8C with 0.033 nmol of tetraubiquitin and increasing quantities of GST-hHR23a: 0.033, 0.067, 0.1, 0.2 and 0.4 nmol, respectively. The samples were treated as described for (a) and probed with anti-ubiquitin (top panel) or anti-GST (bottom panel). As a control, the analogous experiment was performed in the absence of tetraubiquitin to ensure that hHR23a and Rad23 do not interact directly (lane 4 of (b) and (c)). In addition, tetraubiquitin binding to GST-hHR23a (lane 2 of (b)) and His-Rad23 (lane 2 of (c)) but not to either tag or resin (lane 3 of (b) and (c)) was demonstrated. In lanes 5–7 of (d), 20 ml of Ni-NTA resin pre-incubated with 0.1 nmol of His-Ddi1 was mixed for 1 h at 4 8C with 0.132 nmol of tetraubiquitin and increasing quantities of GST-Rad23: 0.1, 0.5 and 1.0 nmol, respectively. The samples were treated as described above and probed with anti-ubiquitin (top panel) and anti-GST (bottom panel). Positive control with no GSTRad23 (lane 2) and a negative control using His-Rad23 UBL with tetraubiquitin (lane 3) were also performed. Without tetraubiquitin, Rad23/Ddi1 interaction under these conditions is weak (lane 4, lower panel), whereas in the presence of tetraubiquitin, the Rad23/Ddi1/tetraubiquitin ternary complex forms (lanes 5–7).

above, tetraubiquitin was found to form a ternary, complex with Rad23 and hHR23a. Furthermore, the addition of hHR23a to resin-bound His-Rad23/ tetraubiquitin reduced the amount of tetraubiquitin that was retained on the resin (Figure 5(c)), a trend that was not observed when Rad23 was added to resin-bound GST-hHR23a/tetraubiquitin (Figure 5(b)). This finding suggests that hHR23a binds tetraubiquitin more strongly than Rad23, which most likely stems from the loss of UBA2mediated homodimerization. Altogether our results support a model in which the Rad23 dimer dissociates upon binding tetraubiquitin and each molecule is used to bind individual moieties. In addition, these experiments suggest that although the hHR23a UBA2 domain

sandwiches between the two ubiquitin moieties of diubiquitin,32 the hHR23a/Rad23 UBA domains can also bind a single ubiquitin moiety of tetraubiquitin, as supported by the complexes that these UBA domains form with monoubiquitin.19,33 Analogously, since monoubiquitin is able to dissociate the Rad23/Ddi1 heterodimer, we hypothesized that these proteins could also form a ternary complex with tetraubiquitin. We therefore incubated tetraubiquitin with Ni-NTA agarose resin containing pre-bound His-Ddi1. After extensive washing, GST-Rad23 was added to the mixture, which was washed extensively again and probed for GST-Rad23 by performing Western blot analysis with anti-GST antibody (Figure 5(d)). This experiment supported our hypothesis that tetraubiquitin

1033

UBL/UBA Proteins Bind a Common Tetraubiquitin

can simultaneously bind two different ubiquitin receptor proteins as much more GST-Rad23 is retained on the resin in its presence.

Discussion Perhaps the most impressive attribute of ubiquitin receptor proteins and polyubiquitin is their adaptability. The ubiquitin moieties of K48-linked tetra-30 and diubiquitin34 pack against each other to form a closed structure. Recent literature, however, suggests a flexible model for polyubiquitin structure,35,36 and diubiquitin forms an open structure to bind the C-terminal UBA domain of hHR23a.32 Here, we demonstrate that tetraubiquitin can bind more than one ubiquitin receptor protein and thereby reveal that it too opens to bind UBA domains. Our rationale for this conclusion is that only the most distal subunit relative to an attached protein substrate is available for binding a UBA domain in its closed structure. The structural changes that occur in polyubiquitin parallel those of Rad23’s human homolog hHR23a. In an earlier study we revealed that hHR23a forms a closed structure via UBL/UBA domain interactions that opens to bind either ubiquitin19 or the proteasome component S5a.29 Here, we demonstrate such UBL/UBA domain interactions to similarly mediate Rad23/Ddi1 heterodimerization and be displaced by ubiquitin. Our findings invite the question: why do UBL/UBA ubiquitin receptor proteins interact? An answer to this question is suggested by two of the results presented here: the Rad23/Ddi1 heterodimer dissociates to bind polyubiquitin and these two proteins can bind a common polyubiquitin chain. Such attributes enable Rad23/Ddi1 heterodimers to bind a greater number of ubiquitin moieties in a polyubiquitin chain than either protein alone. This feature is important because Rad23 binding to polyubiquitin can block its elongation16,24,25 and de-ubiquitylation,25 most likely by precluding chains from the enzymes required for these activities. We therefore propose that dimerization of UBL/UBA domain proteins plays an important role in preventing unnecessary and therefore wasteful chain elongation or premature chain disassembly during the transit of ubiquitylated substrates to the proteasome (Figure 6). In summary, we reveal that UBL/UBA ubiquitin receptor proteins interact with each other in a manner that does not interfere with their binding to polyubiquitin. On the contrary, since these proteins bind a common polyubiquitin chain, we propose such heterodimerization to aid in their ability to bind a greater number of ubiquitin moieties, an attribute that has implications for the chemistry performed on chains during their transit to the proteasome. In humans, Rad23 can bind an additional ubiquitin receptor protein, S5a,20 which contains two ubiquitin interaction motifs (UIMs) rather than UBA domains. The S5a homolog in

Deubiquitylating enzymes

Target

E3 ligases

UBA2

UBA1 UBL

Ub

Ub

Ub

UBA

Ub

Rad23

UBA

UBL

UBL Ddi1 homodimer

26S proteasome

Figure 6. Proposed model for how UBL/UBA domain proteins prevent de-ubiquitylation and unnecessary chain elongation during the transit of substrates to the proteasome. A polyubiquitin chain is greeted with a complex containing multiple UBA domains that are available for binding individual moieties.

yeast (Rpn10) is truncated and lacks the C-terminal UIM to which hHR23a binds.20 Consequently, it does not bind Rad23 and this example of an evolved interaction between two abundant ubiquitin receptor proteins offers further evidence for the importance of their collaborative relationships.

Materials and Methods Sample preparation For NMR spectroscopy Rad23, Ddi1 and single-domain constructs of the Rad23 UBL, UBA1 and UBA2 domains as well as of Ddi1’s UBA domain were each cloned into the pET15b expression vector (Novagen) in-frame with the N-terminal histidine tag. The plasmids containing these genes were each transformed into Escherichia coli BL21 (DE3) cells and grown at 37 8C in M9 minimal medium or in Luria broth containing ampicillin (100 mg/ ml). The cells were harvested 3 h after protein expression was induced with 0.4 mM isopropyl b-D-thiogalactoside (IPTG). The proteins were purified by using affinity purification on Ni-NTA resin as described.21 Further purification was achieved on an FPLC system (Pharmacia), by either Superdex 200 (for full-length Rad23 and Ddi1) or 75 (for UBL and UBA domains) preparative columns. We produced 15N-labeled samples for NMR spectroscopy by growth and expression in M9 minimal medium with [15N]NH4Cl as the only source of nitrogen. Unlabeled monoubiquitin was purchased (SigmaAldrich). NMR spectroscopy All NMR samples were dissolved in 20 mM NaPO4 (pH 6.5), 100 mM NaCl, 0.1% (w/v) NaN3, and 10% 2H2O. Spectra were acquired at 25 8C on Varian NMR spectrometers operating at either 800 MHz or 600 MHz. Processing was performed in NMRPipe37 and the resulting spectra were visualized in XEASY.38 Protein concentrations were calculated by using extinction coefficients based on amino acid composition and absorbance at 280 nm for protein dissolved in 8 M urea.

1034 Western blot analysis GST-tagged Rad23 and hHR23a were produced from the pGEX-2T and pGEX-6P-1 expression vectors (Amersham Pharmacia), respectively, by cloning their cDNA in-frame with glutathione-S-transferase (GST). Each protein was expressed and purified as described29 and Rad23 was separated from GST by cleaving with thrombin. The 0.1 nmol of purified His-tagged Rad23 or Ddi1 or of purified GST-tagged hHR23a was bound to 20 ml of pre-washed Ni-NTA resin or glutathione S-Sepharose resin, respectively. Each resin was allowed to mix at 4 8C overnight with K48-linked tetraubiquitin (Boston Biochem Inc.) and then washed extensively with buffer A (20 mM sodium phosphate (pH 6.5), 100 mM NaCl, 0.5% (v/v) Triton X-100). The Ni-NTA resins were then incubated with either GST-Rad23 or GST-hHR23a for 1 h at 4 8C as the glutathione S-Sepharose resin was mixed with untagged Rad23 under the same conditions. Each resin was pelleted and then washed extensively with buffer A or buffer B (50 mM Tris–HCl (pH 8.0), 100 mM NaCl, 20 mM imidazole, 10% (v/v) glycerol). Proteins that were retained on the resin were fractioned by electrophoresis, transferred to a PVDF membrane, and probed with a polyclonal anti-ubiquitin (Boston Biochem Inc.), anti-Rad23,18 or anti-GST (Santa Cruz Biotechnology) antibody. Visualization was performed using anti-rabbit-horseradish peroxidase and ECL.

Acknowledgements We are grateful to Casey Litchke and Jeannette Zinggeler for assisting with the sample preparation. We also thank Dr Leonard Banaszak for allowing us to use his dynamic light-scattering instrument. NMR data were acquired at the NMR facility of the University of Minnesota and we thank Dr David Live and Dr Beverly Ostrowsky for their technical assistance. NMR instrumentation was provided with funds from the NSF (BIR-961477), the University of Minnesota Medical School, and the Minnesota Medical Foundation. Data processing and visualization were performed in the Basic Sciences Computing Laboratory of the University of Minnesota Supercomputing Institute. This work was funded by grants from the National Institutes of Health CA097004-01A1 (to K.J.W.) and CA099033 (to D.J.C.) as well as by a University of Minnesota Academic Health Center Seed Grant (to K.J.W.) and special grant from the University of Minnesota Cancer Center (to K.J.W. and D.J.C.).

References 1. Ciechanover, A. (1994). The ubiquitin-proteasome proteolytic pathway. Cell, 79, 13–21. 2. Schubert, U., Anton, L. C., Gibbs, J., Norbury, C. C., Yewdell, J. W. & Bennink, J. R. (2000). Rapid degradation of a large fraction of newly synthesized proteins by proteasomes. Nature, 404, 770–774.

UBL/UBA Proteins Bind a Common Tetraubiquitin

3. Rock, K. L. & Goldberg, A. L. (1999). Degradation of cell proteins and the generation of MHC class I-presented peptides. Annu. Rev. Immunol. 17, 739–779. 4. Conaway, R. C., Brower, C. S. & Conaway, J. W. (2002). Emerging roles of ubiquitin in transcription regulation. Science, 296, 1254–1258. 5. Muratani, M. & Tansey, W. P. (2003). How the ubiquitin–proteasome system controls transcription. Nature Rev. Mol. Cell. Biol. 4, 192–201. 6. Yamaguchi, R. & Dutta, A. (2000). Proteasome inhibitors alter the orderly progression of DNA synthesis during S-phase in HeLa cells and lead to replication of DNA. Expt. Cell Res. 261, 271–283. 7. Hicke, L. (2001). A new ticket for entry into budding vesicles-ubiquitin. Cell, 106, 527–530. 8. Katzmann, D. J., Odorizzi, G. & Emr, S. D. (2002). Receptor down regulation and multivesicular-body sorting. Nature Rev. Mol., Cell Biol. 3, 893–905. 9. Hoege, C., Pfander, B., Moldovan, G. L., Pyrowolakis, G. & Jentsch, S. (2002). RAD6-dependent DNA repair is linked to modification of PCNA by ubiquitin and SUMO. Nature, 419, 135–141. 10. Spence, J., Sadis, S., Haas, A. L. & Finley, D. (1995). A ubiquitin mutant with specific defects in DNA repair and multiubiquitination. Mol. Cell. Biol. 15, 1265–1273. 11. Chen, L. & Madura, K. (2002). Rad23 promotes the targeting of proteolytic substrates to the proteasome. Mol. Cell. Biol. 22, 4902–4913. 12. Elsasser, S., Chandler-Militello, D., Muller, B., Hanna, J. & Finley, D. (2004). Rad23 and Rpn10 serve as alternative ubiquitin receptors for the proteasome. J. Biol. Chem. 279, 26817–26822. 13. Kleijnen, M. F., Shih, A. H., Zhou, P., Kumar, S., Soccio, R. E., Kedersha, N. L. et al. (2000). The hPLIC proteins may provide a link between the ubiquitination machinery and the proteasome. Mol. Cell, 6, 409–419. 14. Saeki, Y., Saitoh, A., Toh-e, A. & Yokosawa, H. (2002). Ubiquitin-like proteins and Rpn10 play cooperative roles in ubiquitin-dependent proteolysis. Biochem. Biophys. Res. Commun. 293, 986–992. 15. Kaplun, L., Tzirkin, R., Bakhrat, A., Shabek, N., Ivantsiv, Y. & Raveh, D. (2005). The DNA damageinducible UbL–UbA protein Ddi1 participates in Mec1-mediated degradation of Ho endonuclease. Mol. Cell. Biol. 25, 5355–5362. 16. Verma, R., Oania, R., Graumann, J. & Deshaies, R. J. (2004). Multiubiquitin chain receptors define a layer of substrate selectivity in the ubiquitin–proteasome system. Cell, 118, 99–110. 17. Wilkinson, C. R., Seeger, M., Hartmann-Petersen, R., Stone, M., Wallace, M., Semple, C. & Gordon, C. (2001). Proteins containing the UBA domain are able to bind to multi-ubiquitin chains. Nature Cell Biol. 3, 939–943. 18. Bertolaet, B. L., Clarke, D. J., Wolff, M., Watson, M. H., Henze, M., Divita, G. & Reed, S. I. (2001). UBA domains of DNA damage-inducible proteins interact with ubiquitin. Nature Struct. Biol. 8, 417–422. 19. Wang, Q., Goh, A. M., Howley, P. M. & Walters, K. J. (2003). Ubiquitin recognition by the DNA repair protein hHR23a. Biochemistry, 42, 13529–13535. 20. Hiyama, H., Yokoi, M., Masutani, C., Sugasawa, K., Maekawa, T. et al. (1999). Interaction of hHR23 with S5a. The ubiquitin-like domain of hHR23 mediates interaction with S5a subunit of 26 S proteasome. J. Biol. Chem. 274, 28019–28025.

1035

UBL/UBA Proteins Bind a Common Tetraubiquitin

21. Walters, K. J., Kleijnen, M. F., Goh, A. M., Wagner, G. & Howley, P. M. (2002). Structural studies of the interaction between ubiquitin family proteins and proteasome subunit S5a. Biochemistry, 41, 1767–1777. 22. Elsasser, S., Gali, R. R., Schwickart, M., Larsen, C. N., Leggett, D. S., Muller, B. et al. (2002). Proteasome subunit Rpn1 binds ubiquitin-like protein domains. Nature Cell Biol. 4, 725–730. 23. Saeki, Y., Sone, T., Toh-e, A. & Yokosawa, H. (2002). Identification of ubiquitin-like protein-binding subunits of the 26 S proteasome. Biochem. Biophys. Res. Commun. 296, 813–819. 24. Ortolan, T. G., Tongaonkar, P., Lambertson, D., Chen, L., Schauber, C. & Madura, K. (2000). The DNA repair protein Rad23 is a negative regulator of multi-ubiquitin chain assembly. Nature Cell Biol. 2, 601–608. 25. Raasi, S. & Pickart, C. M. (2003). Rad23 UBA domains inhibit 26 S proteasome-catalyzed proteolysis by sequestering lysine 48-linked polyubiquitin chains. J. Biol. Chem. 278, 8951–8959. 26. Heessen, S., Masucci, M. G. & Dantuma, N. P. (2005). The UBA2 domain functions as an intrinsic stabilization signal that protects Rad23 from proteasomal degradation. Mol. Cell, 18, 225–235. 27. Bertolaet, B. L., Clarke, D. J., Wolff, M., Watson, M. H., Henze, M., Divita, G. & Reed, S. I. (2001). UBA domains mediate protein–protein interactions between two DNA damage-inducible proteins. J. Mol. Biol. 313, 955–963. 28. Rao, H. & Sastry, A. (2002). Recognition of specific ubiquitin conjugates is important for the proteolytic functions of the ubiquitin-associated domain proteins Dsk2 and Rad23. J. Biol. Chem. 277, 11691–11695. 29. Walters, K. J., Lech, P. J., Goh, A. M., Wang, Q. & Howley, P. M. (2003). DNA-repair protein hHR23a

30.

31. 32.

33. 34.

35. 36.

37.

38.

alters its protein structure upon binding proteasomal subunit S5a. Proc. Natl Acad. Sci. USA, 100, 12694–12699. Cook, W. J., Jeffrey, L. C., Kasperek, E. & Pickart, C. M. (1994). Structure of tetraubiquitin shows how multiubiquitin chains can be formed. J. Mol. Biol. 236, 601–609. Thrower, J. S., Hoffman, L., Rechsteiner, M. & Pickart, C. M. (2000). Recognition of the polyubiquitin proteolytic signal. EMBO J. 19, 94–102. Varadan, R., Assfalg, M., Raasi, S., Pickart, C. & Fushman, D. (2005). Structural determinants for selective recognition of a Lys48-linked polyubiquitin chain by a UBA domain. Mol. Cell, 18, 687–698. Mueller, T. D., Kamionka, M. & Feigon, J. (2004). Specificity of the interaction between UBA domains and ubiquitin. J. Biol. Chem. 279, 11926–11936. Cook, W. J., Jeffrey, L. C., Carson, M., Chen, Z. & Pickart, C. M. (1992). Structure of a diubiquitin conjugate and a model for interaction with ubiquitin conjugating enzyme (E2). J. Biol. Chem. 267, 16467–16471. Varadan, R., Walker, O., Pickart, C. & Fushman, D. (2002). Structural properties of polyubiquitin chains in solution. J. Mol. Biol. 324, 637–647. Wang, Q., Young, P. & Walters, K. J. (2005). Structure of S5a bound to monoubiquitin provides a model for polyubiquitin recognition. J. Mol. Biol. 348, 727–739. Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G., Pfeifer, J. & Bax, A. (1995). NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR, 6, 277–293. Bartels, C., Xia, T.-H., Billeter, M., Gu¨ntert, P. & Wu¨thrich, K. (1995). The program XEASY for computer-supported NMR spectral analysis of biological macromolecules. J. Biomol. NMR, 6, 1–10.

Edited by J. Karn (Received 7 October 2005; received in revised form 28 November 2005; accepted 2 December 2005) Available online 19 December 2005